Aletheia 🔍

Aletheia is an AI-powered platform designed to detect, explain, and mitigate irresponsible consumption of information in textual, visual, and video content. It supports researchers, journalists, educators, and the general public in identifying implicit or explicit manipulation in data and responds factually.

---

📌 Objectives

• Provide sentence-level and article-level bias detection for accurate analysis of written content.
• Help rephrase biased statements into neutral alternatives without losing factual accuracy.
• Deepfake detection beyond text to images and videos, acknowledging the multi-modal nature of modern media.
• Store and track analyses securely in Firebase Firestore for research and longitudinal studies.
• Contribute to media literacy and social awareness by making bias measurable, explainable, and correctable.

---

⚙️ Key Features

1. Sentence-Level Bias Detection

• Input: A single sentence.
• Output: Bias classification (neutral, low, medium, high) with a confidence score.
• Provides token-level explanations (which words influenced the bias decision).
• Use case: Researchers studying framing in political speeches or policy documents.

---

2. Article-Level Bias Detection (with Web Scraping)

• Accepts either direct text input or a URL.
• Uses a web scraper to extract article text (removing ads, scripts, and formatting noise).
• Processes the article sentence by sentence, assigning bias levels and generating an overall article bias score.
• Use case: Journalists verifying neutrality in coverage across different media outlets.

---

3. Rephrasing Biased Text

• Powered by the Google Gemini API.
• Rewrites biased content into a more factual and neutral version, preserving key details like names, dates, and figures.
• Both original and rephrased versions are stored in Firestore for reference.
• Use case: Editors and educators ensuring materials are objective and inclusive.

---

4. Image Bias Analysis

• Upload an image to detect potential visual framing bias (e.g., selective cropping, focus, or sentiment conveyed through imagery).
• Provides structured predictions to assist in critical evaluation of visual media.
• Use case: Media analysts assessing bias in photojournalism.

---

5. Video Bias Analysis

• Supports video uploads for asynchronous bias evaluation.
• Returns job IDs for tracking analysis progress.
• Use case: Researchers studying political ads, campaign videos, or news broadcasts.

---

6. Firebase Integration

• Firestore database stores all analyses, including text inputs, rephrased outputs, confidence scores, and timestamps.
• Enables audit trails and long-term studies of bias across sources and time periods.
• Use case: Building datasets for academic research on media bias.

---

🛠 Technology Stack

• Frontend: React
• Backend Framework: FastAPI (Python 3.13), Uvicorn
• ML: HuggingFace Transformers (BERT-based classifiers), SHAP for explanations, CNNs, YOLO, BiLSTM, OpenCV,
• Generative AI: Google Gemini API for rephrasing
• Database: Firebase Firestore
• Task Queue: Celery + Redis for async video processing
• Web Scraping: Requests + BeautifulSoup

---

🚀 Setup Instructions

1. Clone the Repository

git clone https://github.com/your-username/biaslab.git
cd biaslab

2. Create Virtual Environment

python -m venv venv
source venv/bin/activate   # Linux/Mac
venv\Scripts\activate      # Windows

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Variables

Create a .env file inside backend/ with:

GOOGLE_API_KEY=<your-google-api-key>
FIREBASE_CREDENTIALS_JSON=app/secrets/firebase.json

5. Run the Backend API

cd/backend
uvicorn app.main:app --reload

6. Run Frontend

On a separate terminal, run:

cd/frontend
npm install
npm run dev